The evolution of RISS

HP has solidified the position of its HP Storageworks RISS (Reference Information System) in its information lifecycle management (ILM) strategy by turning it into an archive store for all kinds of information: structured (database); unstructured (files); and semi-structured (e-mail). In effect RISS combines three kinds of information, commonly held in different silos, into a single virtual silo - but HP prefers to call them data domains.

RISS was HP's first example of a smart cell-based storage grid. Now RISS has been updated and several new pieces added to the RISS structure.

First of all RISS hardware has changed - the advantages of a commodity approach - with new drives taking capacity up to 1.4TB from 850GB, a useful increment. Secondly there is new compression technology called BSI - standing for Block Single Instance - which uses variable length and fixed length chunking of data making up files. The idea seems to be that a file is broken up into fixed length chunks. Any duplicated chunks are removed and a much smaller pointer is left in their place.

If the file changes then only the changed chunks are stored, with a varying length as necessary, and all other chunks left alone. The effective compression can be increased three to five times by this HP-developed technology. (It might be similar technology to Avamar's 'commonality factoring' in its Axion product.) Previously RISS used file-level single instancing, according to Frank Harbist, HP's VP and General Manager for storage software and ILM.

The combination of increased capacity and hugely-increased storage efficiency means that the effective capacity of the RISS product has possibly quadrupled if not quintupled.

HP's ILM strategy encompasses more than just storage of data and its retrieval, including data capture and data delivery. However the announcements today are focused on information storage and retrieval. What has happened is the original unified RISS product is being divided into parts. The BSI scanning, in which a file's worth of data is separated into single data chunk instances, can be performed by the RISS server itself or by an application connector.

Application connectors
The original application connector was RIM for Messaging, with RIM standing for Reference Information Manager, which focused on Exchange and Notes messages, semi-structured information. RIM for Databases looks at structured information and RIM for Files looks at unstructured information. HP Partners can also use an API to have their applications drive and use RISS.

Harbist contrasts RISS with Centera, saying that: "The RISS archive is content-searchable. Centera doesn't have such a capability. With Centera you need front end (applications) to do this." RISS search can access all data types. Also, from the Compliance point of view it makes obvious sense to have a single archive embracing the three different types of data with a search capability and the ability to verify that data has not been tampered with.

RIM for Databases 2.0 updates version 1.5, which came to HP courtesy of its Outer Bay acquisition. It will now archive older database records into RISS (disk-based) or onto tape and do so in XML format.

BSI capability will be added to RIM for Messaging in the next few months.

HP's own backup software
Data Protector 6.0 is a new version of HP's own backup software, It can now do synthetic backups - long overdue really - and backup to disk, called 'virtual full backup' - a welcome addition. It can carry out instant and automated recovery of Microsoft's Exchange. HP has an Application Recovery Manager product with recovers Exchange and SQL Server. It will be extended to Oracle and makes use of VSS in Windows Server 2003.

If there is a drive failure in an XP or EVA array then the software recovers Exchange or SQL Server, and soon Oracle, to the most recent verifiable snapshot.

But back to Riss; Harbist says: "Data Protector will be able to write eventually direct to RISS." What this implies is that Data Protector becomes another application connector and RISS becomes a general-purpose archive and data protection product.

CDP
Supporting this is the introduction of HP's own continuous data protection (CDP) product, HP Continuous Information Capture Solution. It is based on technology from Mendocino whom HP is partnering. In the future it is likely or probable that the product will feed its CDP information direct into RISS also. RISS is becoming the central component in HP's data protection and archiving strategy.

We have here a 4-level storage concept: fast online disk; secondary disk storage; the RISS disk archive; and HP tape (meaning LTO).

RISS head
RISS is a single level store, but it need not be. The linkage between the RISS software functions and the storage back end will be broken. Harbist talks of a future version with RISS policy management and retrieval functions separated from the back-end storage. HP is thinking of a separate RISS head.

Then it will be possible for RISS to be a multi-level store with its own tiers of disk storage and with tape, even optical disk, included in its back-end storage mix. The RISS head would keep track of what information resided on what tier of storage.

We might even see HP's own stand-alone disk storage virtualiser, the SPS200 - based on HDS' NSC55, recently announced in stand-alone guise - being used to virtualise the storage behind the RISS head.

The announcements today cement RISS as the centrepiece of HP's data protection, archive and compliance strategy.

With its grid capability it is both scalable and resilient. The announcements today provide a convincing reason for customers to re-evaluate RISS and see how its capabilities and direction match up to their own storage strategy.